babble-etl

Data Processing and Analysis Tool Documentation

Overview

This documentation outlines the process of data extraction, cleaning, transformation, and analysis of a CSV file using a configuration file in JSON format.

JSON Configuration Structure

Below is an example of a JSON configuration file structure:

clean_data.json

{
    "data_file": "path/to/your/input.csv",
    
    "columns": [
        "Nest ID", 
        "Nestling ID", 
        "Nestling", 
        "Sex", 
        "Treatment",
    ],
    
    "transformations": {
        "Nest ID": "strip",
        "Date": "to_datetime"
    },
    
    "fill_na": {
        "numeric": 0,
        "string": ""
    },
    
    "rename_columns": {
        "old_name": "new_name"
    },
   
    "data_types": {
        "Nest ID": "str",
        "Nestling ID": "str",
    }
}

Key Fields in the JSON File

• data_file: Path to the raw CSV file to be processed.
• columns: List of columns to be included in the dataset.
• transformations: Dictionary specifying any transformations to be applied to the data.
• fill_na: Dictionary for handling missing values in specific columns.
• rename_columns: Dictionary mapping original column names to new names.
• data_types: Dictionary specifying data types for columns, if applicable.

Command Line Arguments

The following command-line arguments can be used to run the script:

Argument	Type	Required	Default	Description
-i, --input	String	Yes	N/A	Path to the input JSON file for data cleaning and transformation steps.
-m, --minlength	Int	No	2	Minimum length for sequences used in pair analysis.
-k, --kmeans	Int	No	6	Number of clusters for k-means clustering.
-a, --analysis	String	No	N/A	Type of frequency analysis to perform (choices: singles, pairs, triples, quads, quints, all).
-d, --dump	Flag	No	False	Flag to indicate if sequences should be dumped into a plot.
-l, --loglevel	String	No	WARNING	Log level for script execution (choices: DEBUG, INFO, WARNING, ERROR, CRITICAL).
-sc, --sequenceclass	String	No	N/A	Name of Columns in Data Frame to configure the data input for the Sequence Classification Model

Script Workflow

1. Configuration Parsing:
   • The script reads the JSON configuration file specified by the -i argument.
2. Data Extraction:
   • The CSV file specified in data_file is loaded.
3. Data Cleaning and Transformation:
   • The script applies column selection, renaming, transformations, and missing value handling as defined in the JSON configuration.
4. Analysis:
   • Depending on the --analysis argument, the script performs sequence analysis, including singles, pairs, triples, quads, or quints.
5. Logging:
   • Log messages are configured based on the --loglevel argument for better tracking and debugging.
6. Dumping Sequences:
   • If --dump is specified, the processed sequences are saved for further inspection.
7. Sequence Classification:
   • Depending on the --sequenceclass argument, the script performs setting up the df and csv for the Sequence Classification Model

Example Usage

Run the script using the following command:

python script_name.py -i config.json -m 3 -k 5 -a pairs -l INFO --dump --sc "Sex, Treatment"

Explanation:

• Reads configuration from config.json.
• Minimum sequence length is set to 3.
• Runs k-means clustering with 5 clusters.
• Performs pair analysis.
• Logs messages at INFO level.
• Dumps sequences into a plot.
• Makes corresponding df and cvs for Sex and Treatment

Logging

Logs are configured to display timestamps, log levels, and messages to help track the script's progress and troubleshoot any issues.

logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')

Return Value

The script does not return a value but writes logs and, if specified, dumps sequence outputs for plotting.

ANOVA Testing

Analysis of variance (ANOVA) is an extremely important method in exploratory and confirmatory data analysis. ANOVA test can be a valuable tool for exploratory data analysis (EDA) when you want to identify potential differences in the mean of a continuous variable across different categorical groups within your dataset, helping you explore potential relationships and patterns between variables.

Why ANOVA is useful in EDA:

1. Group comparisons: When exploring a dataset, ANOVA can quickly reveal whether there are statistically significant differences between various groups, allowing you to focus your analysis on those areas.

2. Identifying potential relationships: By comparing means across different groups, ANOVA can highlight potential relationships between categorical variables and continuous variables, prompting further investigation.

3. Data exploration: Even if you don't have a specific hypothesis in mind, running an ANOVA can help you discover interesting patterns in your data that might warrant further exploration.

4. Data visualization: The results of an ANOVA can be used to inform visualizations like box plots or bar charts, which can visually represent the differences between groups.

When to Use ANOVA in EDA

• Independent Variable: Categorical (e.g., sex, treatment groups).
• Dependent Variable: Continuous (e.g., vocalization frequency or amplitude).

How to Use ANOVA for EDA

1. One-Way ANOVA: Test differences among means for one categorical independent variable.

   • Example: Do parrot vocalizations differ based on sex (male vs. female)?

2. Two-Way ANOVA: Explore interaction effects between two categorical independent variables.

   • Example: Is there an interaction between sex and treatment on vocalization patterns?

3. Post-Hoc Tests: If the ANOVA reveals significant differences, use post-hoc tests (like Tukey's HSD) to determine which groups differ from each other.

4. Visualization: Complement ANOVA results with visualizations (e.g., boxplots, violin plots) to help interpret differences in group means.